The invention discloses a user mobile game behavior map-based game recommendation method. The method comprises the steps of extracting user characteristics by utilizing a login log of a user in game data; calculating cosine similarity of every two user characteristic vectors, and determining whether two nodes have connected edges or not according to a relationship between a cosine similarity value and a set threshold; extracting operations of the user on a game, such as charging operation of the user on the game and the like, extracting publishing time of the game and time for starting to play the game by the user, and constructing attributes of the user so as to construct a user mobile game behavior map; analyzing the user mobile game behavior map to mine role information of ''fashionable user'' nodes, ''friend'' relationship nodes and the like in the user mobile game behavior map; and endowing neighbor user nodes with different weights according to the role information of the neighbor user nodes connected with a target user node, weighting games which have been played by the neighbor user nodes and games which are not played by the target user node according to the weights to form a recommendation list, and recommending the list to a target user, thereby realizing personalized game recommendation.